ARTICLE pubs.acs.org/OPRD
Multikilogram-Scale Synthesis of a Chiral Cyclopropanol and an Investigation of the Safe Use of Lithium AcetylideEthylene Diamine Complex Ephraim M. Bassan,† Carl A. Baxter,*,‡ Gregory L. Beutner,*,§ Khateeta M. Emerson,*,† Fred J. Fleitz,§ Simon Johnson,‡ Stephen Keen,‡ Mary M. Kim,§ Jeffrey T. Kuethe,§ William R. Leonard,§ Peter R. Mullens,‡ Daniel J. Muzzio,† Christopher Roberge,§ and Nobuyoshi Yasuda§ †
Department of Chemical Process Development and Commercialization, Merck and Co., Inc., Rahway, New Jersey 07065, United States Global Process Chemistry, Merck, Sharp and Dohme Ltd., Hertford Road, Hoddesdon, EN11 9BU, U.K. § Department of Process Research, Merck and Co., Inc., PO Box 2000, Rahway, New Jersey 07065, United States ‡
bS Supporting Information ABSTRACT: A six-step route starting from a readily available vinyl boronate was identified to produce an enantioenriched cyclopropanol in an overall 16% yield. Key steps involve the use of lithium acetylide-ethylene diamine complex 5 and an enzymatic resolution of a racemic cyclopropanol acetate. Process safety considerations surrounding the use of 5 were examined, and an improved procedure is described which was safely demonstrated at multikilogram scale.
’ INTRODUCTION Despite recent advances in asymmetric SimmonsSmith1 and transition metal-catalyzed cyclopropanations,2 the preparation of chiral, nonracemic cyclopropanols remains a formidable challenge.3 The two most common routes to cyclopropanols are the Simmons Smith cyclopropanation of vinyl ethers (Scheme 1, eq 1) and the Kulinkovich cyclopropanation4 (Scheme 1, eq 2). These reactions have both been investigated with chiral catalysts and/or auxiliaries, although the scope and selectivity of these reactions is limited. Shi and co-workers have shown that a dipeptide-derived catalyst gives excellent yields and enantioselectivities in the SimmonsSmith cyclopropanation of ketone-derived silyl enol ethers.5 Corey and co-workers have demonstrated that a titanium TADDOL complex gives moderate levels of enantioselectivity in the Kulinkovich cyclopropanation of an acetate-derived ester.6 Several auxiliary-based methods have also been developed for the SimmonsSmith cyclopropanations of chiral vinyl ethers7 and chiral boronate esters,8 but these methods typically suffer from low yields and difficulties in the removal of the chiral auxiliary. Additionally, the asymmetric synthesis of cyclopropyl boronates has been realized starting from allylic carbonates9 or cyclopropenes,10 but both procedures possess significant limitations in terms of scope. Considering the state of the art for the asymmetric synthesis of cyclopropanols, the development of a scalable synthesis of ent-1 to support an ongoing drug discovery program presented a significant challenge (Scheme 2). Since the asymmetric Simmons Smith cyclopropantion of silyl enol ethers has only been demonstrated with ketone-derived substrates and there is little precedent for the Kulinkovich cyclopropanation of formate esters, the literature provided few options in terms of developing a catalytic asymmetric synthesis for ent-1. Therefore, to access the desired ent-1, an alternative six-step route starting from vinyl boronate 4 was identified which would rely on a final, enzymatic r 2011 American Chemical Society
Scheme 1. Cyclopropanol synthetic methods
Scheme 2. Retrosynthetic analysis of ent-1
resolution to generate the desired, enantioenriched cyclopropanol. The racemic alkynyl cyclopropanol rac-1 could be generated from the corresponding cyclopropanol chloride 2. The cyclopropanol would be unmasked via cleavage of boronate 3 which would be prepared from the E-vinyl boronate 4, a commercially available compound. Even after successful demonstration of this new route at laboratory scale, significant issues remained, particularly in the context of the safe use of lithium acetylideethylene diamine Received: September 7, 2011 Published: December 07, 2011 87
dx.doi.org/10.1021/op2002497 | Org. Process Res. Dev. 2012, 16, 87–95
Organic Process Research & Development
ARTICLE
conditions represented a significant improvement when compared to results using Zn(CH2I)2. With 3 in hand, investigation of the oxidation of the boroncarbon bond was undertaken. Initial attempts to employ sodium perborate8h,14 led to significant decomposition. However, it was found that treatment of 3 with 1 equiv of 10 M sodium hydroxide in methanol at 0 °C followed by 2.0 equiv of 30 wt % hydrogen peroxide gave clean conversion to 2 after an aqueous workup. Initial attempts showed that extractive removal of pinacol was challenging and would require further development for large scale (vide infra); thus, the removal of pinacol from crude 2 was accomplished at this stage of development by column chromatography to give a 69% yield of the desired chloro cyclopropanol 2. At this point, the direct displacement of the primary chloride in 2 to generate the terminal alkyne in 1 was investigated. This would avoid the installation of an oxygen-protecting group and the addition of two steps to the synthesis for protection and deprotection of this early-stage intermediate. The lithium acetylideethylene diamine complex 5 seemed an attractive choice for an alkynylating agent since it is a relatively air-stable, commercially available solid. A survey of the literature on alkynylations with 5 revealed that this reaction was known to proceed even in the presence of protic functionality such as a carboxylic acid.15 However, a more pressing issue was the fact that alkyl chlorides showed poor reactivity with 5. Extensive literature examples exist on the displacement of bromides,16 iodides,17 and oxygen-centered leaving groups such as tosylates, triflates, and epoxides18 with 5, but only a single publication could be found on the reaction of alkyl chlorides.19 A highly polar solvent such as dimethyl sulfoxide (DMSO) was required for the displacement reaction. Investigations on the reaction of 2 with 5 revealed that the only viable solvents for this transformation were N,N0 -dimethylacetamide (DMAc), N-methylpyrrolidinone (NMP), and DMSO. DMAc gave good conversion, but the impurity profile was not as favorable as the reaction performed in either DMSO or NMP. Although the impurities formed during reactions in DMAc were not identified, the possibility of acetylation of either 5 or 1 by DMAc could not be ruled out. The initially developed conditions called for the use of 2.1 equiv of 5 in DMSO (Scheme 6). Due to the low reactivity of the chloride, long reaction times were required at room temperature to obtain >90% conversion. However at 50 °C, after only 1 h, >98% conversion and a 76% yield of rac-1 could be obtained. The crude solution was directly acetylated to give cyclopropyl acetate, rac-7, which could be used directly in the subsequent enzymatic resolution. Although the alkynylation of the unprotected cyclopropanol 2 did avoid the need for a protecting group strategy, the use of excess lithium acetylideethylene diamine complex 5 to perform both the deprotonation of the cyclopropanol and the alkynylation led to the generation of a full equivalent of acetylene. The uncontrolled release of acetylene gas would pose a significant risk upon scale-up. Furthermore, the known incompatibility of DMSO and strongly basic reagents raised additional safety concerns which would require careful investigation prior to implementing this procedure on larger scale.20 An examination of the stability of lithium acetylideethylene diamine complex 5 in DMSO immediately confirmed the concerns about processing the reaction on large scale. Testing was carried out using a closed system accelerating rate calorimeter (ARC) corrected for thermal inertia such that the test results
Scheme 3. Alkynylation of 2 with lithium acetylideethylene diamine complex 5
Scheme 4. Diastereoselective cyclopropanation of 6
complex 5 for installation of the terminal alkyne on larger scale (Scheme 3). In this report, we describe the development and successful implementation of a synthesis of ent-1 on a multikilogram scale. Detailed studies focusing on the hazards of the key alkynylation step with 5 are described which shed light on the hazards of this commercially available reagent and defined a clear path forward for its safe use.
’ RESULTS AND DISCUSSION Development of a Synthesis of the Alkynyl Cyclopropanol rac-1. The cyclopropanation of vinyl boronates has been re-
ported in the literature, predominantly in the context of diastereoselective SimmonsSmith cyclopropanations.8 Initial attempts to perform a diastereoselective cyclopropanation of a tartaramide-modified vinyl boronate 68i gave a 24% yield of 2 after oxidative cleavage, with significant process impurities which could only be removed after a difficult chromatography (Scheme 4). Therefore, the racemic cyclopropanation of the commercially available pinacol vinyl boronate 411 was investigated. Literature precedent existed for this transformation, but only moderate yield was obtained after running the reaction for 10 h at 50 °C.8h In our hands, treatment of 4 with bis-iodomethylzinc (Zn(CH2I)2) prepared from diethylzinc and diiodomethane gave poor conversion (98% conversion to the cyclopropyl boronate 3 in 79% isolated yield after only an acidic aqueous workup (Scheme 5). The high conversion and clean impurity profile obtained under the Shi 88
dx.doi.org/10.1021/op2002497 |Org. Process Res. Dev. 2012, 16, 87–95
Organic Process Research & Development
ARTICLE
Scheme 5. Preparation of chloro cyclopropanol 2
Scheme 6. Alkynylation of the chloro cyclopropanol 2 with 5
Figure 1. Rates of self-heating and pressure generation vs temperature for 5 in DMSO.
Figure 2. Rates of self-heating and pressure generation vs temperature for 5 in DMPU.
mimic what can be expected in the factory (500-L vessel or larger). The test results demonstrated that 5 in DMSO was unstable at temperatures greater than 50 °C, which could potentially lead to thermal runaway (Figure 1). The ARC results showed that around 50 °C a very slow adiabatic runaway reaction started which accelerated to a rapid but controllable rate. A large amount of heat was generated. The approximate adiabatic temperature rise observed was +170 K. This activity self-heated the sample to 210 °C where a second decomposition reaction involving the DMSO solvent occurred, resulting in an extremely violent and uncontrollable runaway reaction. Test results showed that the temperature rose at a rate exceeding 1000 K/min, and an associated pressure generation rate exceeded 100,000 psi/min. Over 4000 psi was generated, causing the test equipment to fail. At this point it was obvious that the process was not safe to scale up. Since initial studies indicated that a polar aprotic solvent would be required for high conversions to rac-1 and DMSO was clearly not suitable for further development, a brief screen of other polar aprotic solvents was undertaken to identify an
alternative (Table 1, Route A). Using 2.1 equiv of lithium acetylideethylene diamine complex 5, 1,3-dimethyl-3,4,5,6tetrahydro-2(1H)pyrimidone (DMPU), DMAc and NMP were identified as potential candidates on the basis of the high conversions observed at room temperature. Examination of the stability of 5 in DMPU using the ARC revealed that 5 in DMPU started to slowly decompose at 70 °C. However, the resulting adiabatic temperature rise was about half of that observed in DMSO, and the maximum rate of temperature increase was reduced by at least an order of magnitude (Figure 2). More importantly, the violent higher-temperature secondary decomposition was not encountered. ARC testing using 5 in NMP gave results comparable to those observed with DMSO for the controllable first exotherm. However, even at temperatures above 215 °C, no secondary decomposition was observed. From this initial data, the thermal stability of 5 in both DMPU and NMP suggested that either would be a safe solvent for scale-up of the reaction of 2 with 5 at 50 °C. However, from a safety perspective, DMPU was clearly the better choice. 89
dx.doi.org/10.1021/op2002497 |Org. Process Res. Dev. 2012, 16, 87–95
Organic Process Research & Development
ARTICLE
Table 1. Survey of solvents for reactions of 2 with 5
route B: 1.1 equiv 5, 1 equiv n-HexLi
route A: 2.2 equiv 5 1 h, RT (% conv)
20 h, RT (% conv)
DMSO
90
>98
DMAc
10
>98
58, 4
24
NMP DMPU
23 >98
>98 ND
98, 2.5 96, 4
56 81
Scheme 7. Optimized conditions for formation of rac-1
conv, time (%, h)
yield (%)
ethylene oxide and butadiene. Due to the special safety controls required for the use of acetylene gas, few if any US pharmaceutical pilot plants are rated for Class 1, Group A. In the EU, however, acetylene has an electrical classification of group IIC, which also includes hydrogen, and some European pharmaceutical pilot plants are rated for group IIC. Although monitoring of the release of acetylene gas during this reaction is required in all cases, the quantification of acetylene in the headspace would have implications regarding where this chemistry could be performed on kilogram scale. In order to understand and quantify the timing and the amount of acetylene produced during this reaction, a ReactIR 4000 was used to measure the evolution of acetylene gas (absorption at 3300 cm1) in the headspace. The ReactIR 4000 was equipped with an online gas cell connected to a 250-mL jacketed flask (see Experimental Section for details). The gas phase in the reactor was examined over the entire course of the reaction by purging dry nitrogen gas through the headspace at a rate of 20 mL/min. A condenser at 5 °C was used to minimize the loss of THF. To quantify the exact amount of acetylene present, calibration with a known mixture of acetylene in nitrogen was used to create a calibration curve. When the optimized procedure using a sacrificial base and 1.1 equiv of 5 (Scheme 7) was performed in the apparatus equipped with the ReactIR, it was observed that acetylene was still evolved during the reaction (Figure 3), with the majority of the acetylene evolving during the early stages of the reaction. It was found that suspending 5 in dry DMPU led to some acetylene release (15 psi.24 The high bond energy of the carboncarbon triple bond makes acetylene explosions more violent than those of most other fuels. For these reasons, the electrical classification of acetylene in the United States is Class 1, Group A.25 Acetylene is categorized by itself in Group A due to the unique hazards of this compound. As a point of reference, Group B includes hydrogen and, in some situations, 90
dx.doi.org/10.1021/op2002497 |Org. Process Res. Dev. 2012, 16, 87–95
Organic Process Research & Development
ARTICLE
Figure 3. ReactIR analysis of reactor headspace for acetylene.
Scheme 8. Enzymatic resolution of rac-7
e0.6% for regular production, placing significant restrictions on the broad utility of this chemistry at plant scale. Despite the low levels of acetylene which can potentially be produced by this reaction, care should be taken when using 5 on any scale to avoid potentially dangerous explosions from the buildup of acetylene. Having examined the safety concerns surrounding lithium acetylideethylene diamine complex 5 and gained a better understanding of the alkynylation of 2, we had some confidence about continuing the development of this route at larger scale. Therefore, with a route to the racemic cyclopropyl acetate rac-7 in hand, investigations of the crucial enzymatic resolution step were undertaken (Scheme 8). Hydrolysis of the cyclopropyl acetate rac-7 was screened against our library of commercially available lipases and amidases, consisting of a number of lyophilized powders, liquid enzyme preparations, and immobilized enzyme preparations. For the lyophilized powders and liquid enzyme preparations the screening was conducted using a 0.1 M, pH = 8.0, aqueous phosphate buffer in a 96-well plate format. The cyclopropyl acetate was dispensed to each reaction tube as a solution in DMSO to give a substrate concentration of 5 mg/mL and a DMSO concentration of 10 vol %. For the immobilized enzyme screen, MTBE was shaken in a separatory funnel with a solution of 0.1 M, pH = 8.0, aqueous phosphate buffer, and then the layers were allowed to separate. The cyclopropyl acetate was then added to a portion of the buffer-saturated MTBE to give a substrate concentration of 5 mg/mL. The substrate/MTBE solution was then mixed with a previously dispensed quantity of the immobilized enzymes in a 96-well plate format at 30 °C. The reactions were then monitored for conversion and enantiomeric excess (see Experimental Section for GC conditions). From the initial screening results, Novozym 435 with MTBE as solvent was selected for further reaction optimization. It was identified that the desired enantiomer ent-1 was the hydrolyzed product of the reaction with rac-7. Screening of additional organic solvents (ethyl acetate, heptane, toluene, THF, acetonitrile, etc.) for an increase in hydrolysis selectivity did not offer any advantages over the initial MTBE conditions. With the MTBE solvent system, no significant background hydrolysis of rac-7 to rac-1 occurred when the MTBE solution was shaken with a 0.1 M aqueous solution of potassium phosphate dibasic (approximate pH of 8.6), indicating that there would be no unselective, uncatalyzed background hydrolysis of this species under these
conditions. Using these more basic conditions removed a pH adjustment from the original procedure and modestly simplified the reaction setup. Finally, the charges of cyclopropyl acetate rac-7 and Novozym 435 were optimized to give a reaction rate that was sufficient to complete the reaction in one day but was still slow enough that overhydrolysis (and a corresponding loss in product enantiomeric excess) could be avoided. When the desired end point of the reaction was obtained (product ee of 96%, approx 40% conversion) the reaction mixture was filtered to remove the enzymatic catalyst and halt any further hydrolysis. The crude reaction mixture was concentrated and then solvent switched to heptane for the subsequent chromatographic purification. Since all of the reactions thus far were to be conducted as a through-process with no isolations or purification, separation of the mixture of ent-1 and rac-7 would serve not only to give the enantioenriched product but also to remove process impurities generated thus far. Purification by column chromatography was successful at removing a variety of impurities which had been formed during this five-step throughprocess and gave material of sufficient quality for the final delivery.
’ LARGE-SCALE DEMONSTRATION OF THE SYNTHESIS OF CYCLOPROPYL ALCOHOL RAC-1 Having established proof of concept for the synthesis of ent-1 as well as identified the relevant issues for the safe handling and use of lithium acetylideethylene diamine complex 5, the procedure was ready for scale-up to multikilogram scale. Although vinyl boronate 4 is commercially available, due to cost and lead-time it was decided to implement a quick synthesis from the chloroalkyne 8 (Scheme 9). The hydroboration of 8 was carried out using pinacol borane in the presence of a catalytic amount of BH3THF (0.05 equiv) and cyclohexene (0.1 equiv).26 The key concerns with this step were the highly exothermic charges of the reagents to an essentially neat reaction mixture. 91
dx.doi.org/10.1021/op2002497 |Org. Process Res. Dev. 2012, 16, 87–95
Organic Process Research & Development
ARTICLE
Scheme 9. Large-scale demonstration of the synthesis of rac-7
The addition of cyclohexene to borane and the subsequent addition of the alkyne 8 to this mixture were highly exothermic, but these could be controlled by adjusting addition rate. A RC-1 (reaction calorimeter) run on the final pinacol borane addition at 30 °C over 2 h showed a 40% accumulation of reagents with a reaction exotherm of 154 kJ/mol and an adiabatic temperature rise of >300 K. Thus, it was decided to do an extended addition over at least 2 h, with careful monitoring of the exotherm throughout the charge. On a 20-kg scale the pinacol borane addition took 4 h 40 min, and after a further 50-min age, the batch temperature stabilized, and 1H NMR showed 85% conversion. After an overnight age, 94% conversion27 was achieved, and the batch was then quenched and worked up to afford 38 kg of 4 in 84% yield. The SimmonsSmith reaction of 4 (80 wt % solution in heptane) was efficient on 20-g runs, achieving >98% conversion after an overnight age. The reactions were very clean, giving 3, after workup and concentration, as a solution in heptane in >90% yield (>95% purity by 1H NMR and GC). Large exotherms were observed upon sequential addition of the reagents; therefore, the reaction was examined further using DSC (differential scanning calorimeter) and RC-1 calorimetry. DSC showed no exotherms in the slurries after additions, quench, and concentration. RC-1 was used for the reagent additions to examine the heat transfer of the reaction in order to calculate a safe addition rate and vessel jacket temperature in the plant. For the specific vessel used in the pilot plant, these data showed the following: (1) Addition rates and vessel jacket temperatures needed to maintain a reaction at 3 °C were: TFA addition to Et2Zn: 32.6 KJ/mol exotherm with adiabatic temperature rise of ∼57 K. TFA addition could be done over 80 min with jacket at 11 °C. CH2I2 addition to TFA/Et2Zn/DCM slurry: 9.3 KJ/mol exotherm with adiabatic temperature rise of ∼12 K. CH2I2 addition could be done over 20 min with jacket at 10 °C. Substrate addition: 23.6 KJ/mol exotherm with adiabatic temperature rise of ∼28 K. Substrate addition could be done over 45 min with jacket at 12 °C. (2) Addition rate and vessel jacket temperature to maintain the quench at 25 °C were: Quench with aqueous HCl: 7.4 kJ/mol exotherm with adiabatic temperature rise of ∼8 K. On scale addition could be done over 10 min with jacket at +6 °C. On scale the SimmonsSmith reaction was run in two batches using 34.3 kg of 4. Both proceeded identically (>98% conversion) and gave 3 as a solution in heptane (44.8 kg at 78 wt %, 96% yield).
For the oxidation of the boronate to the cyclopropyl alcohol, this was achieved by treating boronate 3 with 10 M sodium hydroxide (1.0 equiv) and aqueous hydrogen peroxide (30%, 1.2 equiv). Once the oxidation was complete, the reaction was quenched with hydrochloric acid followed by aqueous sodium sulfite and stirring overnight to ensure complete hydrolysis of any intermediate boronate species. Although all four of the additions are exothermic, running this chemistry in the RC-1 confirmed that there was very little heat accumulation at the end of each of these additions, indicating that the exotherms observed could be controlled by adjusting the addition rate, and that reasonable addition times would be achievable on scale. Furthermore, GC analysis both part way through and at the completion of the peroxide addition indicated that the reaction was almost instantaneous at 5 to 5 °C. In previous lab runs the alcohol 2 was isolated by extraction into MTBE and purification by column chromatography to remove pinacol. It was found, however, that the chromatographic removal of pinacol could be avoided by washing the MTBE extracts with water several times. DCM or toluene could be used in place of MTBE, but its relative ease of removal and disposal made it a better choice for scale-up. Therefore, the quenched reaction mixture was extracted twice with MTBE, and the combined extracts were then washed once with 0.5 M sodium hydroxide and then washed three times with water. The resulting MTBE solution was azeotropically dried by distillation and then concentrated to minimum volume to afford a 75% corrected yield of alcohol 2. The final material typically contained less than 2% pinacol relative to 2 (based on GC) and, excluding solvents, contained only one impurity larger than this (3% LC area percent (LCAP)). A total of 8% of desired product 2 was lost to the initial aqueous layer and the four washes. It was decided to use crude 2 directly in the next step. This process was successfully run on 35 kg scale, without incident, to afford a 27 wt % MTBE solution of alcohol 2, containing 16.1 kg of desired product (83% yield). Pinacol level, aqueous losses, and purity profile all matched typical lab runs. The Li-acetylide/DMPU chemistry developed to avoid the generation of acetylene was scaled. In order to achieve optimum conversion and purity during the course of the reaction when using the bulk stream of 2, the conditions were altered slightly. Upon changing the reaction parameters, we found the most successful and consistent procedure was deprotonation of the alcohol in MTBE and THF at